American Economic Review 2017, 107(10): 3072–3118 https://doi.org/10.1257/aer.20150320 New Evidence on the Aftermath of Financial Crises in Advanced Countries† By Christina D. Romer and David H. Romer* This paper examines the aftermath of postwar financial crises in advanced countries. We construct a new semiannual series on finan- cial distress in 24 OECD countries for the period 1967–2012. The series is based on assessments of the health of countries’ financial systems from a consistent, real-time narrative source, and classifies financial distress on a relatively fine scale. We find that the average decline in output following a financial crisis is statistically signifi- cant and persistent, but only moderate in size. More important, we find that the average decline is sensitive to the specification and sam- ple, and that the aftermath of crises is highly variable across major episodes. A simple forecasting exercise suggests that one important driver of the variation is the severity and persistence of financial dis- tress itself. At the same time, we find little evidence of nonlinearities in the relationship between financial distress and the aftermaths of crises. JEL E32, E44, G01, N10, N20 ( ) Even before the collapse of Lehman Brothers set off a worldwide financial melt- down, economists had shown renewed interest in financial crises. The experiences of Japan and the Nordic countries in the early 1990s and the East Asian crisis of the late 1990s had demonstrated that financial crises were not just a topic of historical interest. Scholars began to examine what previous experiences could tell us about the causes and effects of severe financial disruptions. Not surprisingly, the 2008 crisis added even greater urgency to this research agenda. While researching the aftermaths of crisis is unquestionably important, it is also difficult. Before one can begin to evaluate what typically happens after a crisis, one has to know when crises occurred. But what counts as a “crisis” is far from obvious. * C. Romer: Department of Economics, University of California, Berkeley, Berkeley, CA 94720 email: ( [email protected]; D. Romer: Department of Economics, University of California, Berkeley, Berkeley, ) CA 94720 email: [email protected]. This paper was accepted to the AER under the guidance of John ( ) Leahy, Coeditor. We are grateful to Steven Braun, Jesus Fernandez-Villaverde, Mark Gertler, Pierre-Olivier Gourinchas, Chang-Tai Hsieh, Luc Laeven, Daniel Leigh, Demian Pouzo, James Powell, Carmen Reinhart, Kenneth Rogoff, three anonymous referees, and seminar participants at the University of California, Berkeley, the University of Chicago, Harvard University, Princeton University, Case Western Reserve University, Haverford College, the Banque de France, and the National Bureau of Economic Research for helpful comments and discussions; to Marc Dordal i Carreras, Dmitri Koustas, and Walker Ray for research assistance; and to the Center for Equitable Growth at the University of California, Berkeley for financial support. The authors declare that they have no relevant or material financial interests that relate to the research described in this paper. † Go to https://doi.org/10.1257/aer.20150320 to visit the article page for additional materials and author disclosure statements. ( ) 3072 Vol. 107 no. 10 RomeR and RomeR: the afteRmath of financial cRises 3073 It is natural to want to use a statistical indicator of financial distress, such as an interest rate spread or the change in lending. Such statistical measures have the ben- efits of being objective and capturing variations in the amount of financial disruption across episodes. But, they also have well-known disadvantages. Most obviously, they often do not exist on a consistent basis for a large sample of countries going back in time. More fundamentally, purely statistical indicators may misidentify financial disruptions. For example, financial distress may not show up in an interest rate spread if banks ration credit through quantity restrictions rather than price; or lending may decline because of tight monetary policy or falling output, rather than because of financial disruption. Because of these problems, most studies have taken a different approach to iden- tifying crises. Researchers have combined qualitative evidence from countries’ financial histories with examination of more quantitative measures, such as govern- ment bailouts or bank failures, to date crisis periods. This approach has the ability to capture times of financial distress even when comprehensive statistical indicators do not exist or fail to reflect important financial market disruptions. But, it also has drawbacks. Most such crisis chronologies are a simple binary indicator, and so do not reflect the obvious fact that some crises are worse than others. The measures can also be flawed or difficult to interpret if imprecise criteria for what constitutes a crisis are used, or if they combine diverse phenomena, such as asset price declines, banking problems, and consumer or business bankruptcies. A New Measure of Financial Distress.—In this paper, we create a new semiannual series on financial distress in 24 advanced countries for the period 1967 to 2012. As described in detail in Section I of the paper, our new measure is derived from the contemporaneous narrative accounts of country conditions prepared by analysts at the Organisation for Economic Co-Operation and Development OECD , published ( ) in the OECD Economic Outlook OECD various years . The definition of financial ( ) distress that underlies our new measure is that suggested by Bernanke 1983 : a rise ( ) in the cost of credit intermediation. In this way, we focus on disruptions to credit supply, rather than on broader conceptions of financial problems. We seek to avoid some of the potential drawbacks of more qualitative crisis series by using a precise definition of financial distress, focusing on a single real-time nar- rative source for a large sample of countries over an extended period, and approach- ing the identification as systematically as possible. Like more statistical measures of financial problems, we use the narrative source not merely to identify crisis periods, but to scale the severity of financial distress. Thus we create an indicator that cap- tures the variation in financial disruption across countries and time periods. We find that the new measure identifies many of the same episodes as other crisis indicators. However, some crisis episodes included in other chronologies do not show up in our measure at all. And, the timing of financial distress is often quite different in our new measure than in the existing chronologies. More fundamentally, our scaled measure shows that episodes of financial distress differ greatly in severity and in the way that distress evolves over time. While there are important benefits to relying on a single, consistent, real-time narrative source to identify financial distress, there are also limitations. For exam- ple, the source could be idiosyncratic or biased. We therefore check it against a 3074 THE AMERICAN ECONOMIC REVIEW OCTOBER 2017 wider range of real-time narrative sources. We find that it is reasonably accurate, but certainly not perfect. Moreover, while our series contains information not captured by purely statistical indicators or other crisis chronologies, the fact that it is based only on narrative evidence suggests that those series may also contain information not captured by our series. Thus, each approach is likely to have value added relative to the other. The Average Aftermath of Financial Crises.—In Section II, we use our new measure of financial distress to investigate the typical aftermath of financial crises. Importantly, our narrative source does not provide enough information for us to be able to separate financial distress arising from a decline in output from financial distress occurring for more exogenous reasons. Thus, while our findings provide new evidence on what happens after crises, they contain at most only suggestive evidence of any causal impact of financial distress on real outcomes. To estimate the average or typical aftermath of financial crises, we run straight- forward panel regressions of real GDP on our new measure of financial distress. More specifically, we use the Jordà 2005 local projection method to estimate the ( ) response of GDP at different horizons to an innovation in the financial distress vari- able. We also examine the response of industrial production and unemployment. Consistent with much of the existing literature, we find that in the aftermath of financial crises, real GDP falls significantly and persistently. Importantly, however, for advanced countries in the postwar period, the fall in output following a typical crisis is only moderate. The peak decline in GDP is approximately 6 percent. The fall in industrial production and the rise in the unemployment rate are also statisti- cally significant, but more modest in size. The estimate of the typical aftermath of financial crises using our new measure of financial distress is not dramatically dif- ferent from that derived using existing crisis chronologies for the same time period and sample of countries. When we consider alternative econometric specifications, the estimated nega- tive aftermath of financial crises is noticeably smaller. For example, when we use generalized least squares to take into account the fact that some countries generally have more variable output, the maximum decline in GDP is just 4 percent. Likewise, assuming that the contemporaneous relationship between financial distress and out- put reflects the effect of output on distress, and so should not be included as part of the aftermath, reduces the estimated negative outcome by a similar amount. Variation in the Aftermath of Financial Crises.—In some ways, focusing on the average aftermath of crises obscures the more fundamental issue of the variation in aftermaths. Section III explores this topic. We show that particular episodes are important outliers, and that excluding them changes the estimated average response substantially. For example, when we split the sample in 2007, the estimates for the earlier sample are somewhat smaller than those for the sample that includes the 2008 global financial crisis. Similarly, the decline in GDP in Greece following the 2008 crisis was so large that simply excluding Greece from the sample lowers the estimated average output decline following a crisis by more than a percentage point. The second way that we investigate the variation in aftermaths is through a sim- ple forecasting exercise. Focusing on episodes of significant financial distress, we Vol. 107 no. 10 RomeR and RomeR: the afteRmath of financial cRises 3075 compare an autoregressive univariate forecast conditional on information up through the year before high distress with the actual behavior of GDP. We find substantial dif- ferences in these forecast residuals across episodes. For example, GDP fell little rela- tive to its precrisis path following the financial crises in Norway and the United States in the early 1990s, but dramatically following the crises in Japan in the 1990s and Turkey in the early 2000s. Such differences are even more pronounced in the 2008 episode, when many advanced countries suffered significant distress. Some coun- tries, such as Norway, Austria, and the United States, show relatively small forecast residuals, while others, such as Iceland, Spain, and Greece, show very large ones. We go on to investigate the role that the severity and persistence of financial distress may play in accounting for the variation in aftermaths across episodes of high distress. We expand the simple forecasting framework to include the actual evolution of distress throughout each episode. This analysis shows that between a third and half of the variance of the univariate forecast error in these episodes can be accounted for by differences in financial distress itself. We also investigate the possibility that the response of output to financial dis- tress is nonlinear in the severity of distress. For example, perhaps extreme levels of distress have a disproportionately large effect on economic activity. To do this, we consider nonlinear permutations of the Jordà approach. In no instance are the non- linearities large in either an economic or statistical sense. Related Work.—As described above, a large modern literature has developed on the identification and aftermath of financial crises. Caprio and Klingebiel 1996, ( 1999, 2003 did pioneering work on deriving a crisis chronology for a wide range of ) countries. Their crisis chronology is based in part on the retrospective assessments of experts on financial developments in various countries. Kaminsky and Reinhart 1999 is an early study comparing the behavior of output and other variables before ( ) and after the start of crises, compared with averages in “tranquil” times. Bordo et al. 2001 refine the Caprio and Klingebiel chronology, and also provide early esti- ( ) mates of the impact of crises. Reinhart and Rogoff, in their influential book, This Time Is Different 2009a , ( ) and a number of related papers see, for example, 2009b, 2014 , also derive a cri- ( ) sis chronology, based in part on earlier studies. Using their chronology and a wide range of outcome measures, they find important commonalities in both the run-up to crises and their aftermaths. In recent years, scholars at the International Monetary Fund IMF have refined the Caprio and Klingebiel dates using more precise criteria ( ) and some quantitative indicators see Laeven and Valencia 2014 for the most recent ( description of the IMF chronology . Recent work by Krishnamurthy and Muir ) 2016 investigates credit spreads as a possible indicator of financial disturbances, ( ) and finds a substantial correlation between this statistical measure of financial dis- tress and common crisis chronologies. Studies have investigated the behavior of the real economy following financial crises in a variety of ways. Reinhart and Rogoff 2009a look at the peak-to-trough ( ) fall in output per capita around crises. Bordo et al. 2001 ; IMF 2009a ; Schularick ( ) ( ) and Taylor 2012 ; Jordà, Schularick, and Taylor 2013 ; and Claessens, Kose, and ( ) ( ) Terrones 2014 not only examine recessions around financial crises, but explic- ( ) itly compare recessions with and without crises. These studies find that recessions 3076 THE AMERICAN ECONOMIC REVIEW OCTOBER 2017 accompanied by financial crises are more severe. Similarly, Claessens, Kose, and Terrones 2009 compare recessions with and without “credit crunches,” where ( ) credit crunches are identified based on the magnitudes in the declines in credit. Hoggarth, Reis, and Saporta 2002 , IMF 2009b , and Laeven and Valencia 2014 ( ) ( ) ( ) compare the path of output following crises with projections of precrisis trends. These studies find that output often falls far below the precrisis path, but that there is substantial dispersion across episodes. A few studies use standard regression analysis of postwar data. Cerra and Saxena 2008 look at the behavior of output following the starting dates of the banking ( ) crises identified by Caprio and Klingebiel 2003 . They find large and persistent ( ) falls in output after the onset of crises. Gourinchas and Obstfeld 2012 , combining ( ) dates of banking crises from a range of existing chronologies, estimate updated ver- sions of regressions analogous to the averages reported by Kaminsky and Reinhart 1999 .1 ( ) Most studies consider banking crisis in samples that combine advanced and other countries. A few studies, such as Cerra and Saxena 2009 , IMF 2009b , Gourinchas ( ) ( ) and Obstfeld 2012 , and Claessens, Kose, and Terrones 2009, 2014 , report results ( ) ( ) for advanced or high-income countries separately. In general, these studies find that though the aftermaths of financial crises are less severe in advanced countries, they are still quite poor. Schularick and Taylor 2012 and Jordà, Schularick, and Taylor ( ) 2013 look just at a sample of advanced countries, but over a very long sample ( ) period. They find substantial declines in output following crises, and also that the size of the credit boom preceding crises is an important predictor of the size of the decline. I. New Measure of Financial Distress The central contribution of this study is the derivation of a new scaled measure of financial distress for 24 advanced countries for the period 1967–2012. A. Approach Definition of Financial Distress.—Conceptually, we think of financial distress as corresponding to increases in what Bernanke 1983 calls the “cost of credit ( ) intermediation.” This cost includes both the cost of funds for financial institutions relative to a safe interest rate, and their costs of screening, monitoring, and adminis- tering loans and other types of financing. A rise in the cost of intermediation makes it more costly for financial institutions to extend loans to firms and households, and thus reduces the supply of credit. Importantly, we do not consider reductions in lending stemming from increases in all interest rates as a result of tighter monetary ( 1 A study that is similar to ours in approach, but that focuses only on the United States, is Jalil 2015. Jalil con- ( ) structs a new series on banking panics for the United States back to the early 1800s using contemporary newspaper accounts. He scales panics into major and minor crises, and identifies a handful of panics that appear to have been caused by factors other than a decline in output. Using simple time-series regressions, he finds that crises have large and persistent real effects in the period before 1929. A study that focuses on the United States over both the prewar and postwar periods using more traditional business-cycle analysis is Bordo and Haubrich 2017. They find that ( ) recoveries following financial crises are not slower than other recoveries. Vol. 107 no. 10 RomeR and RomeR: the afteRmath of financial cRises 3077 policy, for example as representing financial distress. The question of how mone- ) tary policy and the overall level of interest rates affect the economy is different from the issue of the aftermath of disruptions to the financial system, and we do not want to confound the two.2 Narrative Evidence.—Following most previous work, we do not rely on statis- tical indicators of financial distress. Rather, we rely on more qualitative evidence about the health of the financial system to construct our index of financial distress. In particular, we use a careful analysis of a single, real-time narrative source to deduce times when the cost of credit intermediation rose. The use of contemporane- ous accounts should help us avoid the natural tendency to perhaps look a little harder for a financial crisis before a known severe recession, or to identify the start of a crisis earlier than was apparent in real time. The use of a single source that covers many countries over a long period of time helps ensure consistency in the analysis across countries and episodes. A second important feature of our measure is that we do not treat financial crises as a 0–1 variable, or divide crises into just two groups, such as minor and major or nonsystemic and systemic. Both logic and descriptions of actual episodes of finan- cial distress suggest that financial-market problems come much closer to falling along a continuum than to being discrete events that are all of similar severities, or that fall into just a few categories. Treating a continuous variable as discrete intro- duces measurement error, both because the variation across crises is omitted and because a small inaccuracy in evaluating an observation can cause a large change in the value assigned to it. Source.—The particular real-time narrative source we use is the OECD Economic Outlook. This is a semiannual publication that describes economic conditions in each member country of the OECD at mid-year and year-end. The volumes have been published since 1967. This source has several advantages. First, and most obviously, it is relatively high frequency, available over a long time period, and covers a large number of advanced countries. Thus it allows us to construct a measure of distress for a large sample over much of the postwar period. Second, the entries are analytical and of medium length a typical entry is roughly 2,000 words . As a result, they provide serious ( ) information in a relatively concise form. Third, the format, topics covered, and level of analysis appear to be relatively consistent both across countries and over time. Thus, the source can be used to derive a measure of financial distress for a number of countries that is similarly consistent across countries and time. Finally, financial conditions and determinants of credit growth are discussed routinely in the volumes from the beginning of the sample, and bank health is often mentioned. As a result, financial distress is likely to be captured if it is present. 2 Bernanke also includes influences on credit flows and interest rates resulting from changes in the creditworthi- ness of borrowers in his definition of the cost of credit intermediation. Because our goal is to examine the aftermath of financial distress and because considering the creditworthiness of borrowers blurs the line between loan supply and loan demand, we focus only on the condition of financial firms. 3078 THE AMERICAN ECONOMIC REVIEW OCTOBER 2017 To have a relatively consistent sample and to keep the focus on advanced coun- tries, we restrict the sample to the 24 members of the OECD as of 1973.3 Given that the OECD Economic Outlook begins in 1967, that is the starting date of our analysis. We go through the second half of 2012, so that we capture the 2008 finan- cial crisis. B. Implementation Methods.—To derive our new scaled measure of financial distress, we read the OECD Economic Outlook to see if OECD analysts described a rise in the cost of credit intermediation for individual countries. We put the most weight on factors that are clear markers for increases in the cost of intermediation. We look for discus- sions of such developments as increases in financial institutions’ costs of obtaining funds relative to a safe interest rate; general increases in the perceived riskiness of financial institutions; reductions in financial institutions’ willingness to lend; dis- ruptions in normal borrower-lender relationships that make it harder for financial institutions to evaluate prospective borrowers; and difficulties of creditworthy bor- rowers in obtaining funds because of problems at financial institutions. In addition to looking for descriptions of factors directly linked to the cost of intermediation, we look for references to developments likely to weaken financial institutions, and so reduce their ability to perform their normal functions. Examples include rising loan defaults, increases in nonperforming loans, balance sheet prob- lems, and erosion of banks’ capital. To scale the degree of financial distress, we attempt to group episodes that the OECD Economic Outlook describes in similar terms together, and to place ones that it describes as more severe in higher categories. In this grouping and ordering, we look for signs of more or less change in the indicators mentioned above. Was the rise in the perceived riskiness of financial institutions relatively minor, or so large that it is described as a widespread panic? Was the effect on the willingness to lend described as minor or extreme? Was the rise in nonperforming loans thought to be small or large? We also consider some indirect proxies for the size of the rise in the cost of inter- mediation. For example, we put some weight on descriptions of government inter- vention in the financial system as an indicator of the perceived severity of balance sheet and funding problems. However, we do not use this information mechanically. We try to take into account the fact that aggressive government intervention, rather than indicating a large rise in the cost of intermediation, might prevent any signif- icant rise; or that greatly delayed intervention might clean up institutions that had long since become insolvent and whose lending activities had already been super- seded by healthier institutions. Likewise, we tend to use discussions of widespread bank failures as an imperfect indicator of a more severe loss of confidence in finan- cial institutions, and hence of a more dramatic increase in the cost of credit inter- mediation. We again try to be cognizant of the fact that institutions’ cost of credit 3 The countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Luxembourg, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, the United Kingdom, and the United States. Vol. 107 no. 10 RomeR and RomeR: the afteRmath of financial cRises 3079 intermediation, and hence their ability to lend, can change greatly without their outright failure—particularly in the presence of regulatory forbearance, or of just enough government intervention to prevent outright failure. Finally, the OECD’s descriptions of the actual or anticipated impact of financial troubles on spending and the economy are often a useful summary indicator for the perceived severity of financial distress.4 Criteria for the Different Categories.—The categories to which we assign episodes have natural interpretations. Our main ones are credit disruption, minor crisis, moderate crisis, major crisis, and extreme crisis. In keeping with the fact that the accounts suggest that financial-market problems fall along continuum, we sub- divide each category into regular, minus, and plus. Thus, for example, an episode of relatively minor financial distress could be classified as credit disruption–minus, credit disruption–regular, or credit disruption–plus. In our empirical work, we con- vert these categories into a numerical scale. Cases where there is no financial distress are assigned a zero. Positive levels of distress start at 1 for a credit disruption–minus and go through 15 for an extreme crisis–plus. The hallmark of the episodes that we identify as credit disruptions is that the OECD perceived strains in financial markets, funding problems, or other indicators of an increase in the cost of credit intermediation that were important enough to be mentioned, but that it did not believe were having significant macroeconomic consequences. A common form for this to take was for the OECD to describe the problems not as directly affecting its outlook for the country, but as posing a risk to the outlook. Other possibilities are that the OECD viewed the problems as affecting only a narrow part of the economy; that it mentioned them in passing or explicitly identified them as minor; or that it described the financial system as improved but not fully healed following a situation that we classify as a minor crisis. An example of a regular credit disruption occurred in Germany in 1974:2 that is, the second half ( of 1974 , where the OECD described “strains” in the banking system and the exten- ) sion of special credit facilities to help small and medium-sized companies obtain credit OECD 1974:2, pp. 50 and 26, respectively . ( ) A canonical case of a minor crisis has three characteristics: a perception by the OECD that there were significant problems in the financial sector; a belief that they were affecting credit supply or the overall performance of the economy in a way that was clearly nontrivial, and not confined to a minor part of the economy; and a belief that they were not so severe that they were central to recent macroeconomic developments or to the economy’s prospects. An example of a regular minor crisis is France in 1996:1, where the OECD described significant problems in the banking sector, including “high refinancing … costs and large provisions for bad debts,” as well as government intervention to support some financial institutions, but did not give banking problems a central role in its discussion of the outlook OECD 1996:1, ( p. 78 . ) A moderate crisis, in our classification, involves problems in the financial sec- tor that are widespread and severe, central to the performance of the economy as 4 Importantly, we see no evidence in the OECD Economic Outlook that OECD analysts were deducing financial distress from declines in spending and output. Rather, they viewed distress as one influence on those outcomes. 3080 THE AMERICAN ECONOMIC REVIEW OCTOBER 2017 a whole, and not so serious that they could reasonably be described as the finan- cial system seizing up entirely. One specific criterion we use is whether the OECD mentioned the financial-sector problems prominently—for example, in the open- ing summary of the entry on a country. Another is whether the OECD discussed impacts on credit supply or real activity repeatedly. We also take descriptions of sizeable government interventions in the financial system as an indicator of a mod- erate crisis. Thus, our definition of a moderate crisis represents a quite significant level of financial distress, and appears to roughly correspond to the cutoff in other chronologies, such as Caprio et al. 2005 , Reinhart and Rogoff 2014 , and Laeven ( ) ( ) and Valencia 2014 , between a crisis and no crisis, or between a systemic crisis ( ) and a nonsystemic crisis. An example of a regular moderate crisis is Sweden in 1993:1, where the OECD Economic Outlook referred to “an unprecedented increase in banks’ loan losses” and “the capital bases of most major banks rapidly eroding,” and described substantial government rescue operations OECD 1993:1, p. 115 . It ( ) also said, “greater weakness of demand could be accentuated by rising capital costs in the event of larger loan losses” OECD 1993:1, p. 115 . ( ) At the severe end of the spectrum are major and extreme financial crises. These are situations where there are large impediments to normal financial intermediation throughout virtually all of the financial system. In identifying these episodes, we look for such markers as the unreserved use of the term “crisis” in referring to the financial system, and for such terms as “dire,” grave,” “unsound,” and “paralysis.” We also look for clear-cut statements that the financial-sector disruptions were hav- ing an important effect on credit supply and macroeconomic outcomes. In addition, we view references to major government interventions as suggesting that the prob- lems were severe. We find only a handful of major and extreme crises in our sample. An example is Japan in 1998:2, which we classify as an extreme crisis–minus. In that case, the OECD referred to the “breakdown in the credit creation mechanism,” to “the severe and prolonged crisis in the banking system,” and to banks being in “dire straits” OECD 1998:2, pp. 44, 20, and 45, respectively . ( ) Our subdivision of the broad categories into minor, regular, and plus is based on the specifics of the discussions within these general rubrics. In the case of credit disruptions, for example, we tend to place disruptions that the OECD described as posing moderate risks to the outlook in higher categories than ones that it viewed as posing minor risks. Similarly, if the OECD reported that a disruption was serious enough that it had caused authorities to make some type of intervention in credit markets to improve credit flows, we tend to classify the disruption as more serious. Documentation.—Online Appendix A provides more information about our crite- ria for the different categories of financial distress and our procedures for classifying episodes using the accounts in the OECD Economic Outlook. Table 1 lists each epi- sode for which we identify a positive level of financial distress. The bulk of online Appendix A provides episode-by-episode explanations of the analysis and discussion in the OECD Economic Outlook that lead to our classifications. Thus, it should enable others to check our interpretation and classification of the narrative accounts. Exhibit 1 reproduces the Appendix entries for the four episodes cited above: Germany in 1974:2 credit disruption–regular , France in 1996:1 minor crisis–regular , Sweden ( ) ( ) in 1993:1 moderate crisis–regular , and Japan in 1998:2 extreme crisis–minus . ( ) ( ) Vol. 107 no. 10 RomeR and RomeR: the afteRmath of financial cRises 3081 Table 1—Financial Distress in OECD Countries, 1967:1–2012:2 Australia France Continued Italy Continued ( ) ( ) 2008:1 Credit disrupt.–reg. 2011:2 Minor crisis–reg. 2010:1 Minor crisis–plus 2008:2 Minor crisis–minus 2012:1 Minor crisis–minus 2010:2 Minor crisis–minus 2009:1 Credit disrupt.–minus 2012:2 Credit disrupt.–minus 2011:1 Credit disrupt.–reg. 2011:2 Minor crisis–reg. Austria Germany 2012:1 Moderate crisis–minus 2008:2 Moderate crisis–minus 1974:2 Credit disrupt.–reg. 2012:2 Moderate crisis–reg. 2009:1 Moderate crisis–minus 2003:1 Credit disrupt.–minus 2009:2 Minor crisis–minus 2007:2 Credit disrupt.–reg. Japan 2010:1 Minor crisis–minus 2008:1 Minor crisis–minus 1990:2 Credit disrupt.–plus 2010:2 Credit disrupt.–minus 2008:2 Minor crisis–plus 1991:1 Credit disrupt.–minus 2011:1 Credit disrupt.–minus 2009:1 Minor crisis–reg. 1991:2 Minor crisis–reg. 2011:2 Credit disrupt.–plus 2009:2 Minor crisis–minus 1992:1 Credit disrupt.–plus 2012:1 Minor crisis–reg. 2010:1 Minor crisis–minus 1992:2 Minor crisis–minus 2012:2 Credit disrupt.–plus 2010:2 Credit disrupt.–reg. 1993:1 Minor crisis–minus 2011:1 Credit disrupt.–minus 1993:2 Minor crisis–reg. Belgium 2011:2 Credit disrupt.–plus 1994:1 Credit disrupt.–plus 2008:2 Minor crisis–minus 2012:1 Credit disrupt.–reg. 1994:2 Credit disrupt.–plus 2009:1 Credit disrupt.–plus 2012:2 Credit disrupt.–reg. 1995:1 Minor crisis–minus 2009:2 Credit disrupt.–reg. 1995:2 Minor crisis–reg. 2011:2 Credit disrupt.–reg. Greece 1996:1 Minor crisis–plus 2008:2 Minor crisis–minus 1996:2 Minor crisis–minus Canada 2009:1 Moderate crisis–minus 1997:1 Minor crisis–reg. 2007:2 Credit disrupt.–plus 2009:2 Minor crisis–plus 1997:2 Moderate crisis–minus 2008:1 Minor crisis–minus 2010:1 Moderate crisis–reg. 1998:1 Major crisis–reg. 2008:2 Minor crisis–reg. 2010:2 Moderate crisis–minus 1998:2 Extreme crisis–minus 2009:1 Minor crisis–reg. 2011:1 Moderate crisis–minus 1999:1 Moderate crisis–plus 2009:2 Credit disrupt.–plus 2011:2 Moderate crisis–minus 1999:2 Minor crisis–plus 2012:1 Moderate crisis–reg. 2000:1 Minor crisis–minus Denmark 2012:2 Moderate crisis–reg. 2000:2 Credit disrupt.–plus 2008:1 Credit disrupt.–minus 2001:1 Minor crisis–plus 2008:2 Minor crisis–plus Iceland 2001:2 Minor crisis–plus 2009:1 Moderate crisis–minus 2006:2 Minor crisis–reg. 2002:1 Moderate crisis–reg. 2009:2 Minor crisis–plus 2007:1 Credit disrupt.–reg. 2002:2 Moderate crisis–minus 2010:1 Credit disrupt.–plus 2007:2 Credit disrupt.–plus 2003:1 Minor crisis–plus 2011:2 Credit disrupt.–plus 2008:1 Moderate crisis–reg. 2003:2 Minor crisis–reg. 2012:1 Credit disrupt.–reg. 2008:2 Major crisis–reg. 2004:1 Minor crisis–minus 2012:2 Credit disrupt.–minus 2009:1 Extreme crisis–reg. 2004:2 Credit disrupt.–plus 2009:2 Moderate crisis–plus 2005:1 Credit disrupt.–reg. Finland 2010:1 Moderate crisis–minus 2008:2 Credit disrupt.–plus 1992:1 Credit disrupt.–reg. 2010:2 Moderate crisis–minus 2009:1 Minor crisis–minus 1992:2 Minor crisis–plus 2011:1 Moderate crisis–minus 2009:2 Credit disrupt.–plus 1993:1 Moderate crisis–reg. 2011:2 Minor crisis–minus 2010:1 Credit disrupt.–minus 1993:2 Minor crisis–reg. 2012:1 Credit disrupt.–reg. 1994:1 Credit disrupt.–plus Luxembourg 2008:2 Minor crisis–minus Ireland 2008:1 Credit disrupt.–reg. 2009:1 Minor crisis–minus 2007:2 Credit disrupt.–minus 2008:2 Minor crisis–reg. 2009:2 Credit disrupt.–minus 2008:2 Minor crisis–plus 2009:1 Credit disrupt.–plus 2009:1 Moderate crisis–minus 2009:2 Credit disrupt.–reg. France 2009:2 Minor crisis–plus 2010:1 Credit disrupt.–minus 1991:2 Credit disrupt.–minus 2010:1 Moderate crisis–reg. 2011:2 Credit disrupt.–reg. 1995:1 Credit disrupt.–reg. 2010:2 Moderate crisis–minus 1995:2 Minor crisis–minus 2011:1 Moderate crisis–reg. Netherlands 1996:1 Minor crisis–reg. 2011:2 Minor crisis–plus 2008:1 Credit disrupt.–minus 1996:2 Minor crisis–reg. 2012:1 Minor crisis–reg. 2008:2 Credit disrupt.–plus 1997:1 Credit disrupt.–plus 2012:2 Minor crisis–plus 2009:1 Minor crisis–minus 2007:2 Credit disrupt.–plus 2009:2 Credit disrupt.–minus 2008:1 Credit disrupt.–reg. Italy 2011:2 Credit disrupt.–minus 2008:2 Moderate crisis–minus 1997:1 Credit disrupt.–minus 2012:1 Minor crisis–minus 2009:1 Minor crisis–plus 2007:2 Credit disrupt.–minus 2012:2 Credit disrupt.–reg. 2009:2 Minor crisis–minus 2008:1 Minor crisis–minus 2010:1 Credit disrupt.–plus 2008:2 Moderate crisis–minus New Zealand 2010:2 Credit disrupt.–minus 2009:1 Minor crisis–plus 2007:2 Credit disrupt.–reg. 2011:1 Credit disrupt.–minus 2009:2 Minor crisis–reg. 2008:1 Credit disrupt.–plus Continued ( )